M-Vet Livestock Dataset

M-Vet Livestock Dataset is an open-access dataset created by the M-Vet project (www.m-vet.net) based at Makerere University Artificial Intelligence & Data Science Research Lab (www.air.ug) and supported by Lacuna Fund. It is aimed at supporting machine learning models for image classification tasks in Livestock. This dataset contains about 18,000 images of different animal types, including cows, goats, and pigs, collected from various farms and regions and annotated for animal type with corresponding classes. The dataset is designed to facilitate research and development in livestock management, particularly in animal classification tasks using computer vision. It is a valuable resource for researchers, developers, and agricultural stakeholders looking to innovate in animal health monitoring and diagnostics. Available on GitHub for public use.

The dataset consists of nine subfolders (0001 to 0009), each containing three directories: labels, images, and data. Each image has a corresponding .txt file containing its annotations.

For example, given the image:

• M-Vet_Livestock-Dataset-main/0001/images/e0b206bf-ee2f-4d6a-bdbe-ea29d70402aa7725714119516389484_jpg.rf.31cf1c70bbd1a46bfb83404ffe9414dc.jpg
• The corresponding label file: M-Vet_Livestock-Dataset-main/0001/labels/e0b206bf-ee2f-4d6a-bdbe-ea29d70402aa7725714119516389484_jpg.rf.31cf1c70bbd1a46bfb83404ffe9414dc.txt
• Contains the following annotation: 0 1 0.07107843124999999 0.311274509375 0.07107843124999999 0.311274509375 0.51992034375 1 0.51992034375 1 0.07107843124999999

Acknowledgement: Dataset is created by M-Vet project(www.m-vet) led by Daniel Mutembesa(linkedin.com), in collaboration with 162 Expert and rural based Veterinarians and a network of over 1,500 Livestock farmers in Uganda, the National Livestock Resources and Research Institute(naro.go.ug), Veterinarians Without Boarders (vetswithoutbordersus.org), and Research Consortium on African Swine Fever at Makerere University.